Re: ATM Robo-foucalt...is it calibrated?

["Bob May" <bobmay@nethere.com>]

As a person who has written code to do automated calibration of equipemtn,
one thing I will say is that automation of the process makes it a lot more
reliable for results.  The process may be wrong but it will be wrong each
and every time it is done and in the same degree and fashion.  In other
words, it may be wrong but every one of the sets of answers is going to be
exactly the same wrong answer.
This said, the process that is being done with the RoboFoucault program is
going to tend to want to produce correct answers.  The first thing is that
the light detection is looking for a midpoint which is a hard thing to do
incorrectly, especially when that midpoint requirement is a broad one to
match - 50% errors from that midpoint only degrade the quality of the
measurement otherwise by a very small number.
The matching of the brightness levels between the two sides is done to
within 1% (actually about 1 count over about 200 counts of brightness of the
pixels in the image) and any bias will tend to be a consistent error and
thus will be removed from the answer as being of little consequence.  If
more precision is needed then it is possible to bracket the range and derive
the slope of the difference and thus you can gain a huge amount of accuracy
by prediction of the slope.
All of the basic setup errors (tester motions not pointed at the mirror,
etc.) of the test are going to be the same as for a hand done setup of the
test.  The thing here is that the automated tester can complain about the
lack of alignment if the KE needs to be moved more than a certain amount
while doing the set of measurements.  In addition, if necessary, length
measurements can be recalculated because of the angle error and the
corrected values provided for further work.
All in all, I'd tend to prefer to do the automated test for its reliability.
Bob May
http://nav.to/bobmay
bobmay@nethere.com
NEW! http://bobmay.astronomy.net